Multivariate differential association analysis

Area of Science:

• * Statistical genetics
• * Bioinformatics
• * Computational biology

Background:

• * Understanding how relationships between variables change across conditions is crucial in scientific research.
• * Comparing biological systems often involves examining differences in genomic feature relationships between cases and controls.

Purpose of the Study:

• * To evaluate whether dependence relationships between two sets of high-dimensional variables differ across two distinct conditions.
• * To develop a novel statistical test for detecting differential dependence.

Main Methods:

• * Proposed a new kernel-based test to assess the similarity of dependence relationships under two conditions.
• * Introduced an asymptotic permutation null distribution for the test statistic.
• * Demonstrated computational efficiency for large-scale data analysis.

Main Results:

• * The proposed test effectively captures differential dependence between variable sets.
• * Numerical studies confirmed high power in detecting both linear and non-linear differential relationships.
• * The method proved reliable in finite sample scenarios.

Conclusions:

• * The new kernel-based test provides a powerful and efficient tool for identifying differential dependence across conditions.
• * The kerDAA R package facilitates the application of this method in practical research.
• * This approach enhances the analysis of complex relationships in large datasets.